---

AI Mail Support Assistant

An intelligent, AI-powered communication assistant designed to streamline customer support operations. This tool automatically fetches, categorizes, prioritizes, and generates context-aware draft responses for support emails, significantly reducing manual workload and improving response times.

Demo

Watch the demo video here: YouTube Link

Architecture & Approach

Documentation: Google Document

Features

Core Functionality

• Automated Email Retrieval: Fetches incoming emails from a Gmail account via IMAP, filtering for support-related keywords (support, query, request, help).
• AI-Powered Analysis: Utilizes Google's Gemini AI to perform:
  • Sentiment Analysis: Classifies emails as Positive, Negative, or Neutral.
  • Priority Scoring: Flags emails as URGENT or NOT_URGENT based on keywords and sentiment.
  • Information Extraction: Parses key details like contact info, order IDs, and specific requests into a structured JSON format.
  • Context-Aware Draft Generation: Creates professional, empathetic, and knowledgeable draft responses using a built-in knowledge base (RAG).
• Prioritized Processing: Implements a priority queue, ensuring urgent emails are processed and displayed first.

Dashboard & User Interface

• Unified Dashboard: A clean, modern UI built with React and Material-UI.
• Email Inbox: Lists all processed emails with sender, subject, sentiment, and priority status.
• Detailed Email View: Click any email to see its full content, extracted information, and the AI-generated draft response.
• Interactive Analytics: Displays key metrics and charts, including:
  • Total, pending, and resolved email counts.
  • Emails received in the last 24 hours.
  • Breakdown of emails by sentiment and priority.
• Search Functionality: Quickly find emails by sender, subject, or body content.

Tech Stack

Backend

• Runtime: Node.js
• Framework: Express.js
• Database: PostgreSQL with Prisma ORM
• Email Protocol: IMAP (using imap-simple and mailparser)
• AI/LLM: Google Gemini AI API

Frontend

• Library: React 18 with TypeScript
• UI Framework: Material-UI (MUI)
• HTTP Client: Axios
• Charts: Recharts

Project Structure

ai-mail-support/
├── client/                 # React TypeScript Frontend
│   ├── public/
│   ├── src/
│   │   ├── components/     # Reusable UI components (Dashboard, Stats, EmailList, etc.)
│   │   ├── api.ts         # Axios configuration
│   │   ├── theme.ts       # MUI theme customization
│   │   └── types.ts       # TypeScript type definitions
│   ├── package.json
│   └── ...
├── server/                 # Node.js Express Backend
│   ├── services/
│   │   ├── emailService.js # IMAP fetching and processing logic
│   │   └── aiService.js    # Gemini AI integration & prompt engineering
│   ├── prisma/
│   │   └── schema.prisma  # Database schema
│   ├── index.js           # Express server setup and API routes
│   ├── package.json
│   └── ...
└── README.md

Getting Started

Prerequisites

Before you begin, ensure you have the following installed on your system:

• Node.js (v18 or higher)
• npm or yarn
• PostgreSQL database
• A Gmail account (with Less Secure App Access enabled or an App Password)

1. Clone the Repository

git clone <your-repo-url>
cd ai-mail-support

2. Backend Setup

1. Navigate to the server directory:

   cd server

2. Install dependencies:

   npm install

3. Set up Environment Variables: Create a .env file in the server/ directory and configure the following variables:

   DATABASE_URL="postgresql://username:password@localhost:5432/ai_mail_support"
   GEMINI_API_KEY="your_google_gemini_api_key_here"
   IMAP_USER="[email protected]"
   IMAP_PASSWORD="your_gmail_app_password"
   PORT=3001

   How to get a Gmail App Password: Gmail App Passwords

4. Set up the Database: Run Prisma migrations to create the tables in your PostgreSQL database.

   npx prisma generate
   npx prisma db push

5. Start the backend server:

   npm run dev

   The server will start on http://localhost:3001. It will automatically begin fetching and processing emails every 10 minutes.

3. Frontend Setup

1. Open a new terminal and navigate to the client directory:

   cd ../client

2. Install dependencies:

   npm install

3. Start the React development server:

   npm start

   The client will start on http://localhost:3000 and should automatically connect to the backend.

How to Use

1. View the Dashboard: Open your browser to http://localhost:3000. The dashboard will load with statistics and your inbox.
2. Wait for Emails: The server checks for new emails every 10 minutes. Send a test email to your configured Gmail address with a subject containing "Support", "Help", etc.
3. Analyze Results: New emails will appear in the inbox list, categorized by sentiment and priority. Urgent emails will appear at the top.
4. View Details & Drafts: Click on any email to open a detailed view. Here you can see the original email, the extracted information, and the AI-generated draft response.
5. Copy Drafts: Use the copy icon next to the draft response to copy it to your clipboard for use in your email client.

API Endpoints

Endpoint	Method	Description
/api/emails	GET	Fetches all processed emails, sorted by priority (urgent first) and date.
/api/stats	GET	Returns dashboard statistics (total counts, sentiment/priority breakdown, 24h volume).

Database Schema (Prisma)

The application uses a PostgreSQL database with a single Email model storing all processed data.

model Email {
  id            String   @id @default(cuid())
  messageId     String   @unique
  sender        String
  subject       String
  body          String
  receivedAt    DateTime
  sentiment     Sentiment
  priority      Priority
  extractedInfo Json?    // Stores names, phone numbers, etc.
  draftResponse String?
  status        Status
  createdAt     DateTime @default(now())
  updatedAt     DateTime @updatedAt
}

AI & Prompt Engineering

The core intelligence is in server/services/aiService.js. It uses a carefully engineered prompt to instruct the Gemini model to:

1. Analyze sentiment and priority based on defined rules.
2. Extract structured information from unstructured text.
3. Generate a empathetic and professional draft response by grounding its knowledge in the provided knowledgeBase string (a simple RAG implementation).

Potential Improvements & Future Work

• Email Sending: Integrate with SMTP or the Gmail API to send responses directly from the dashboard.
• Response Editing & Approval: Allow users to edit and approve drafts before sending.
• Knowledge Base Management: Create a UI to manage the knowledge base content dynamically.
• Real-time Updates: Implement WebSockets for real-time email notifications instead of polling.
• Multiple Email Providers: Add support for Outlook, Yahoo, and other providers via their APIs.
• User Authentication: Add login/logout functionality to support multiple agents.
• Advanced Analytics: Add more detailed charts and time-series analysis for support performance.

License

This project was created as part of a hackathon.

---